Research On Daily Atmospheric Visibility Estimation Based On Digital Image

In view of the important impact of visibility on daily production and life,this paper focuses on the research of the method of obtaining visibility information based on digital images.Compared with the shortcomings of the traditional visibility visual method,such as the disadvantages of high input cost and complicated erection requirements,the visibility method based on video image has the advantages of low input cost,simple erection operation and wide deployment range.It is a new idea of visibility measurement and has broad application prospects in the future.At the same time,this paper adopts the advanced dark channel prior principle to conduct the estimation research of visibility.Compared with previous research results,the program has the advantage of not having to set up the target observation point and the operation is simple.The main research contents of this paper are summarized as follows:(1)Based on the dark channel prior principle and the principle of visibility measurement,the principle formula for estimating visibility using the dark channel prior principle is derived.For the optimization problem of the rough estimated transmittance,the effects of the soft matting method and the guided filtering method are compared in detail.It is found that the guiding filter has obvious superiority in the calculation efficiency of the transmittance when the processing effect is not much different.At the same time,the method of replacing the color image by grayscale image for guiding filtering's guide image is adopted,and the operation efficiency of the guiding filtering is further improved,and the improvement is up to 95.93%.(2)The influence of the fog concentration coefficient on visibility estimation and the main factors affecting the selection of fog concentration coefficient are discussed.Fog concentration coefficient is the key factor for the transmittance correction.The image gray mean value and weather conditions are important factors affecting the selection of fog concentration coefficient.Therefore,this paper achieves the adaptive selection of fog concentration coefficient by constructing BP neural network and completing the training,and achieves good results.(3)The observational experiment was carried out in 10 days from May 7 to May 16,2018 by using the self-adaptive fog concentration coefficient estimation system.The results of the visibility estimation were compared with those of the Scattering Visibility meter.It was found that the experimental results met the requirements of 20% visibility error,with the accuracy of 97.1% and the error of 2.9%.(4)Based on the hypothetical model of atmospheric radiation theory,a new method for calculating visibility using dark channel priori principle is derived.This method can get rid of the limitation of the information of the target scene and is more convenient to use.Comparing with the Scattering Visibility meter,the measurement results basically meet the error requirements,and the accuracy is over 80%,which is slightly lower than the traditional scheme.Moreover,its calculation amount is large,and the real-time performance is not enough.Finally,this paper discusses the deficiencies in the research,the possible causes and solutions,and further proposes the research focus and direction.